What money is for, and the three jobs it has to do

You have money sitting in an account. Someone tells you it's melting. Someone else tells you cash is a position. A third person tells you the dollar is doomed and hands you a chart. None of them start by telling you what money is, so none of the arguments actually land.

Here's the version that matters to your office. Money isn't one thing you either trust or don't. It's three separate jobs, and a currency can be excellent at two of them while quietly failing the third. If you can't name the three jobs, you can't tell which one is breaking, and you'll either panic early or notice late.

Start with no money at all. You have a skill, someone else has bread, and you want to trade. This is barter, direct swap of goods for goods, no money in the middle.

Barter has one killing flaw. Economists call it the double coincidence of wants: for a trade to happen, you have to want what I have and I have to want what you have, at the same time, in the same place, in the same amounts. Miss any of those and there's no deal.

William Stanley Jevons, a British economist, opened his 1875 book Money and the Mechanism of Exchange with a true story that makes this physical. A French singer, Mademoiselle Zélie, toured the world and performed in the Society Islands for a third of the ticket receipts. Her fee came in as three pigs, twenty-three turkeys, forty-four chickens, five thousand coconuts, plus bananas, lemons and oranges. Back at the Halle market in Paris that haul would have fetched around four thousand francs, decent pay for five songs. In the islands it was a farm she had to feed with the fruit until it could be eaten or given away.

She got paid. She just didn't get paid in money. Money is whatever solves that problem, and it solves it by doing three jobs at once.

Job one, medium of exchange: the thing everyone accepts in trade, so you never need a double coincidence. You sell your work for it, you buy anything with it.

Job two, unit of account: the yardstick you quote prices and keep books in. Economists call it the numéraire, which just means the one unit everything else is measured against. The Bank for International Settlements, the bank that central banks bank with, describes the unit of account in its 2026 Annual Economic Report as the unit value is denominated in, and stresses that what matters is the collective agreement to use the same one for economic calculation.

Job three, store of value: it holds purchasing power between the day you earn it and the day you spend it. Not forever. Over your holding period.

Three jobs. One word. That's the whole confusion.

Now the machine, step by step.

Step one: search costs. In pure barter, every trade needs a match. Finding matches costs time, and time is the real price. So people start holding goods they don't personally want, because those goods are easy to pass on. Economists call that quality salability, how quickly and cheaply you can resell something near its going value.

Step two: one good pulls ahead. Salability feeds on itself. If half the camp takes cigarettes, taking cigarettes is smart even if you don't smoke, which makes it three quarters, which makes it everyone. This is a network effect, where the value of a thing rises with the number of people using it. Money is the oldest network effect there is.

We have a clean experiment. R. A. Radford was a British economist held as a prisoner of war, and in November 1945 he published "The Economic Organisation of a P.O.W. Camp" in the journal Economica. Prisoners got Red Cross parcels, started with barter, and converged on cigarettes as currency. Prices got posted on notice boards. A camp paper currency was later introduced and failed. When cigarettes ran short, the men fell back to barter and the little economy got visibly worse, because there was no longer a single thing everyone would take.

Step three: the price list collapses. This is the underrated part. Under barter, every good has a price in every other good. With 100 goods that's 4,950 separate exchange rates to know, which is just the arithmetic of pairing 100 things two at a time. Pick one good as the yardstick and you have 100 prices. That's the unit of account job arriving, and it's a compression of information, not a convenience. Cheap comparison is what lets you notice that one supplier is charging more than another, which is what lets competition work at all.

Step four: contracts stretch across time. Once prices are quoted in one unit, you can write agreements in it. Wages, rent, a loan, a lease, an insurance policy. Every one of those is a promise about a future amount of the unit. Now the third job becomes load-bearing, because both sides of the contract are betting the unit still means something on the payment date.

Step five: money becomes something you don't have to think about. The BIS calls good money an information-insensitive asset, meaning you accept it without investigating the issuer. You take a twenty without running a credit check on anybody. That property rests on what the BIS calls singleness: claims in the unit redeem at par, one for one, with central bank money, with finality. Your bank deposit, the cash in your pocket and the balance in a payment app are all "a dollar" because they swap one for one on demand. The BIS describes the standard arrangement as two-tier: private banks lend and innovate, the central bank supplies the final settlement asset and sets the rules.

Step six: the material is optional, the memory isn't. Milton Friedman's 1991 Hoover working paper "The Island of Stone Money" is five pages that rewire how you see this. On Yap, in the Caroline Islands, huge carved stones served as money. The stones mostly didn't move. Ownership changed by common knowledge, and Friedman notes that when German colonial officials wanted the footpaths repaired, they simply painted black crosses on stones to mark them as claimed, and the paths got repaired. Nothing physical changed. Friedman then puts the mirror up: in 1932 and 1933 the Bank of France, worried the US might leave the gold standard, asked the New York Fed to convert dollars to gold and just relabel it in the vault. The gold never crossed the ocean, and the press ran headlines about American gold losses. Money is a shared ledger of who owns what, and the coin, note or stone is just the format the ledger happens to use.

Step seven: the jobs fail in order. This is the part to keep. Store of value goes first, quietly, because it's the only job measured over years. Unit of account goes second: contracts shorten, prices get re-quoted weekly, then daily, then in a foreign currency while payment still happens in the local one. Medium of exchange goes last, because habit and law are sticky, and it's the job people notice, which is why by the time it's obvious the other two are long gone.

The feedback loop is mechanical. If holding the unit is a loss, you spend it faster. Faster spending is more money chasing the same goods, so prices rise, which confirms that holding is a loss. The BIS's own working paper on money and trust points out the flip side for volatile assets: swings undermine a thing's use as unit of account and means of payment, which is why prices stay quoted in national currencies even where other assets trade heavily.

Step eight: so you split the jobs on purpose. Nothing says one asset must do all three. Cash is superb at exchange and accounting and mediocre at storage. Long-duration assets are the reverse. Almost every real portfolio is an answer to the question of which thing does which job, whether or not the owner ever phrased it that way.

Jevons wrote the modern framing. His 1875 book runs Barter, then Exchange, then a chapter titled "The Functions of Money," and that chapter list is why every textbook since teaches the same three or four jobs. Central banks still teach it that way: the Federal Reserve Bank of Philadelphia's classroom lesson on the functions and characteristics of money and the Federal Reserve's own "Functions of Money" resource run the same track for students.

Then the field tests. Yap, a German colony from 1899 to 1919, showed money working with the object nailed down and only the ledger moving. Radford's camp in 1945 showed money emerging from scratch in weeks among people who'd lost everything else. Both point the same way: the three jobs get filled by whatever's available, and a community will invent a currency rather than live without one.

The failure case is Zimbabwe. Official inflation reporting stopped after July 2008, with the last published figure at 231,150,888.87%. Steve Hanke of the Cato Institute and Johns Hopkins reconstructed the rest from market exchange rates, and puts the peak in November 2008 at 89.7 sextillion percent, that's 89.7 followed by 22 zeros, the second worst hyperinflation ever recorded. Hanke also corrects the widely repeated IMF number, putting September 2008 at 471 billion percent rather than 500 billion. Notice what happened first, though: long before the notes were refused, Zimbabweans were quoting prices in US dollars. The unit of account went before the medium of exchange did. And this isn't ancient history. Writing in October 2017, Hanke logged Zimbabwe hyperinflating again at 348%, the 58th verified episode on the Hanke-Krus table of world hyperinflations.

The live chapter is digital. The BIS devoted a full chapter of its June 2026 Annual Economic Report to stablecoins, and it tests them against exactly these properties, a common unit and redemption at par with finality, rather than against whether they're new.

Five figures worth carrying.

Jevons, 1875: one singer's fee came to three pigs, twenty-three turkeys, forty-four chickens and five thousand coconuts, a haul worth roughly four thousand francs in Paris and worth a feeding problem in the Society Islands. That's the cost of no money, priced in livestock.

Arithmetic, mine, not sourced to anyone: 100 goods under barter imply 4,950 exchange rates, and 100 under a single unit of account imply 100 prices. The unit of account is a fifty to one cut in what you have to know.

Zimbabwe, July 2008: the last official inflation print, 231,150,888.87%, after which the statistics agency stopped publishing. Zimbabwe, November 2008: Hanke's reconstructed peak of 89.7 sextillion percent. Zimbabwe, October 2017: 348% and back on the hyperinflation table, the 58th verified episode.

Radford, November 1945: Economica, volume 12, number 48, pages 189 to 201. Thirteen pages, written by a man who'd just been released, still the best short account of money being born.

For the slow version of the same story in your own currency, FRED publishes the purchasing power of the consumer dollar as series CUUR0000SA0R. Pull it yourself. The shape of the line is the lesson, and Lesson 1.4 takes it apart.

An office doesn't ask "is the dollar good money." It asks which job each account is doing, and whether the tool matches the job.

Cash. Cash held for buying is doing the medium-of-exchange job, and it's supposed to be boring. Judge it on availability and settlement speed, not on return. Losing a few percent a year on money you'll deploy inside twelve months is the fee for being able to act on a Wednesday.

Plan. The office picks one unit of account and keeps score in it, permanently. Usually that's the currency your rent, taxes and groceries are priced in, because those are the obligations you actually have to meet. Switching yardsticks mid-year is how people talk themselves into believing a flat year was a good one.

Positioning. Store of value is a separate job with a separate horizon, and it's the only job where the answer changes over decades. The office's frame is duration matching: money you need in a year sits in the exchange bucket, money you don't need for ten sits in the storage bucket, and the two don't get confused because they're literally in different accounts.

Structure. Watch the order of failure, because it's an early-warning system. Contracts getting shorter, prices being re-quoted more often, a second currency creeping into quotes: those are unit-of-account symptoms, and they show up years before anyone refuses the notes.

Intel, never advice. You make the moves.

Our paper fund keeps score in dollars because that's what it spends. Cash sits until there's a reason to use it, and it's never graded on yield. Everything else is there to do the store-of-value job over years, so it gets judged over years. When we review a holding, the first question is which of the three jobs it was hired to do.